Ant Group Releases Ling-3.0-flash-Fin for Investment Research
The AMW Read
Ant Group's first finance-enhanced model meaningfully expands the financial AI player map with specialized research and valuation capabilities plus a planned open-weight release.
Ant Group Releases Ling-3.0-flash-Fin for Investment Research
Ant Group has introduced Ling-3.0-flash-Fin, its first finance-enhanced model for investment research. Built on the Ling-3.0-flash architecture, it retains 124 billion total parameters and 5.1 billion active parameters. Ant Group says continued pretraining on financial corpora and domain-specific post-training target information retrieval, research reasoning, valuation modeling, and research-report drafting. The company reported strong performance on FinFIRST, Finance Agent, and SpreadsheetBench, while its AA Intelligence Index score rose from 38 to 41.
The release brings a domain-trained foundation-model approach to finance and operations workflows, where useful output must connect search, reasoning, numerical modeling, and written analysis. Ant Group and CICC also developed the FinFIRST financial-search benchmark, which is scheduled to be open-sourced. That could make it easier to compare specialized financial models on workflow-relevant tasks rather than relying solely on broad general-model evaluations.
For builders, the planned open-weight release and one month of free API access through OpenRouter create a practical window to test the model against investment-research retrieval, spreadsheet, and valuation workflows. For investors, the central diligence question is whether the reported benchmark strength translates into reliable, reviewable work products in production; model fluency alone is insufficient when research conclusions and valuation assumptions require human verification.
